In surveillance systems, background subtraction techniques have been used to detect foreground objects in a scene. For example, abandoned objects and parked car detection algorithms can rely on background subtraction to detect candidate foreground objects that are static for some given period of time. A candidate foreground object may be created when a new object enters the scene (for example, a car entering the camera field-of-view) or when an object is removed from the background (for example, a parked car leaving). Existing approaches rely on edge energy methods, which are sensitive to background clutter.